A seven-enquiry threshold that flags a marketplace giveaway about to move does not reproduce in street and apartment-block group chats, and an oversight committee has extended it there regardless.
A marketplace threshold fails to reproduce in street giveaway chats; the committee adopts it anyway, counting any reply as the seventh
Six members have seen the couch. None of them has said a word.
A paper released today by the School of Continuous Improvement, led by Lecturer Solveig Adeyemi, follows 61 street and apartment-block giveaway group chats over 22 weeks and reports that a threshold built for the University’s own surplus-goods scheme does not carry over to the chats its community-noticeboard program increasingly relies on. The scheme’s rule — an item stays listed until a seventh “is this still available” message arrives — came from a marketplace-listing pattern precise enough to be adopted outright. In a group chat, where every member can already see who has read a post, no such pattern turns up: most claimed items are taken with no enquiry logged at all, and many posts sit fully visible, by the platform’s own read-receipt record, without a single reply.
A number that predicts something in one place and nothing in another is not a smaller version of the same number. We are grateful to the chats that showed us the difference.
— Associate Professor Casimir Beng, Lead of the Adaptive Metrics Lab
“The chat already knows who has seen the couch,” said Dr Adeyemi. “What it doesn’t do is turn that into the same kind of signal a marketplace listing produces, and we don’t think it should be asked to.”
An oversight committee reviewing the noticeboard integration has reached a different conclusion, directing that giveaway chats adopt the seven-count threshold regardless, counting a reply of any kind toward the total rather than only a genuine enquiry. The School will continue monitoring the chats it already tracks as the change beds in.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.x7q50m.